| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | |
| 4 | from unittest.mock import Mock |
| 5 | |
| 6 | import pytest |
| 7 | |
| 8 | import tests.utils as test_utils |
| 9 | from tests.utils import RemoteLaunchRenderServer, RemoteOpenAIServer |
| 10 | |
| 11 | |
| 12 | @pytest.mark.parametrize( |
| 13 | ("args", "expected"), |
| 14 | [ |
| 15 | (["--api-key", "key1", "key2"], ["--api-key", "***", "***"]), |
| 16 | (["--api_key", "key1", "key2"], ["--api_key", "***", "***"]), |
| 17 | (["--api-key=key1", "key2"], ["--api-key=***", "***"]), |
| 18 | (["--api_key=key1", "key2"], ["--api_key=***", "***"]), |
| 19 | (["--hf-token", "token"], ["--hf-token", "***"]), |
| 20 | (["--hf_token", "token"], ["--hf_token", "***"]), |
| 21 | (["--hf-token=token"], ["--hf-token=***"]), |
| 22 | (["--hf_token=token"], ["--hf_token=***"]), |
| 23 | ], |
| 24 | ) |
| 25 | def test_redact_sensitive_cli_arg_variants(args: list[str], expected: list[str]): |
| 26 | assert test_utils._redact_sensitive_cli_args(args) == expected |
| 27 | |
| 28 | |
| 29 | @pytest.mark.parametrize( |
| 30 | ("server_cls", "command"), |
| 31 | [ |
| 32 | (RemoteOpenAIServer, ["vllm", "serve"]), |
| 33 | (RemoteLaunchRenderServer, ["vllm", "launch", "render"]), |
| 34 | ], |
| 35 | ) |
| 36 | def test_server_redacts_sensitive_values( |
| 37 | server_cls, |
| 38 | command: list[str], |
| 39 | monkeypatch: pytest.MonkeyPatch, |
| 40 | capsys: pytest.CaptureFixture[str], |
| 41 | ): |
| 42 | server = object.__new__(server_cls) |
| 43 | popen = Mock() |
| 44 | monkeypatch.setattr(test_utils.subprocess, "Popen", popen) |
| 45 | monkeypatch.setenv("INHERITED_SECRET", "inherited-secret") |
| 46 | serve_args = [ |
| 47 | "--api_key", |
| 48 | "api-secret-1", |
| 49 | "api-secret-2", |
| 50 | "--hf-token=hf-secret", |
| 51 | "--max-num-seqs", |
| 52 | "2", |
| 53 | ] |
| 54 | |
| 55 | server._start_server( |
| 56 | "test-model", serve_args, {"OVERRIDE_SECRET": "override-secret"} |
| 57 | ) |
| 58 | |
| 59 | stdout = capsys.readouterr().out |
| 60 | assert "api-secret-1" not in stdout |
| 61 | assert "api-secret-2" not in stdout |
| 62 | assert "hf-secret" not in stdout |
| 63 | assert "inherited-secret" not in stdout |
| 64 | assert "override-secret" not in stdout |
| 65 | assert "--api_key *** ***" in stdout |
| 66 | assert "--hf-token=***" in stdout |
| 67 | assert "--max-num-seqs 2" in stdout |
| 68 | assert popen.call_args.args[0] == [*command, "test-model", *serve_args] |
| 69 | child_env = popen.call_args.kwargs["env"] |
| 70 | assert child_env["INHERITED_SECRET"] == "inherited-secret" |
| 71 | assert child_env["OVERRIDE_SECRET"] == "override-secret" |
| 72 | |
| 73 | |
| 74 | def test_openai_server_shutdown_wait_covers_engine_cleanup( |
| 75 | monkeypatch: pytest.MonkeyPatch, |
| 76 | ): |
| 77 | server = object.__new__(RemoteOpenAIServer) |
| 78 | server._request_shutdown_timeout = 0.0 |
| 79 | server.proc = Mock(pid=1234) |
| 80 | engine_timeout_args = [] |
| 81 | |
| 82 | def get_engine_timeout(request_timeout, process_timeout): |
| 83 | engine_timeout_args.append((request_timeout, process_timeout)) |
| 84 | return 60.0 |
| 85 | |
| 86 | monkeypatch.setattr( |
| 87 | test_utils, "get_engine_process_shutdown_timeout", get_engine_timeout |
| 88 | ) |
| 89 | monkeypatch.setattr(test_utils.os, "getpgid", lambda pid: pid) |
| 90 | monkeypatch.setattr(server, "_kill_process_group_survivors", Mock()) |
| 91 | |
| 92 | server._terminate_process_tree() |
| 93 | |
| 94 | assert engine_timeout_args == [(0.0, 0.0)] |
| 95 | server.proc.wait.assert_called_once_with(timeout=75.0) |